Current input method in helm-mini

[Dmitry Cherkassov]

I've noticed that current-input-method in helm-mini is the same as in current buffer. (I guess some update changed that).

Can i haz it back (so they are independent) ?

Thanks.

[Thierry Volpiatto]

Dmitry Cherkassov <dcher...@gmail.com> writes:

> I've noticed that current-input-method in helm-mini is the same as in
> current buffer. (I guess some update changed that).

Yes, it is generally what people using current-input-method expect.

> Can i haz it back (so they are independent) ?

Just use `C-\' (toggle-input-method) once you are in helm.

See https://github.com/emacs-helm/helm/issues/797

--
Thierry
Get my Gnupg key:
gpg --keyserver pgp.mit.edu --recv-keys 59F29997

[Eric Abrahamsen]

Thierry Volpiatto

<thierry....@gmail.com> writes:

> Dmitry Cherkassov <dcher...@gmail.com> writes:
>
>> I've noticed that current-input-method in helm-mini is the same as in
>> current buffer. (I guess some update changed that).
>
> Yes, it is generally what people using current-input-method expect.
>
>> Can i haz it back (so they are independent) ?
>
> Just use `C-\' (toggle-input-method) once you are in helm.
>
> See https://github.com/emacs-helm/helm/issues/797

But it is pretty frustrating when you're typing in one language, and
then run helm-M-x, and then have to switch languages to type in the
command name (then switch back again). I appreciate that this whole area
is pretty complicated, but is there any more user customization that can
be provided here? I don't have any commands named in Chinese!

Thanks

[Thierry Volpiatto]

I personally don't use input-methods.
The people who sent this issue wanted this behavior, so I did, now some
other people want something else.
I can't satisfy everybody, sorry, I think what we have actually is
reasonable, if you are using chinese in your current buffer, expect
chinese in your minibuffer, up to you to turn this off (which is not a
big deal).

[Eric Abrahamsen]

Thierry Volpiatto

Fair enough :)

[Michael Heerdegen]

Eric Abrahamsen <er...@ericabrahamsen.net>
writes:

> Fair enough :)

FWIW, I think what Eric describes is really annoying. Setting the input
method for the minibuffer should be done in a hook, so that users can
decide to keep or remove it.

I know you (Thierry) want to keep things simple, and adding an option
for that would indeed be overkill. But to enable users to configure
such stuff via a hook seems better to me than hardcoding one behavior.


Regards,

Michael.

[Thierry Volpiatto]

Michael Heerdegen <michael_...@web.de> writes:

> and adding an option for that would indeed be overkill. But to enable
> users to configure such stuff via a hook seems better to me than
> hardcoding one behavior.

If you really insist, done.
I don't see what a hook have to do with this though.

[Michael Heerdegen]

Thierry Volpiatto

<thierry....@gmail.com> writes:

> I don't see what a hook have to do with this though.

My idea of a fix was different. We have this line in
`helm-initial-setup':

(helm-set-local-variable 'current-input-method current-input-method)

Make it a defun and put it into some hook (`helm-before-initialize-hook'
would probably work). Then the user could control input method
inheritance by adding/removing the new function to/from the hook.

This would spare us adding a new option.